| Tue Gorgens, University of New South Wales |
| Semiparametric Estimation of Single-Index Transition Intensities |
| Session: C-12-19 Wednesday 16 August 2000 by Gorgens, Tue |
| This research develops semiparametric kernel-based estimators of state-specific conditional transition intensities, h(y|x), for duration models with right-censoring and/or multiple destinations (competing risks). Both discrete and continuous duration data are considered. The maintained assumption is that h(y|x) depends on x only through an index x'b. In contrast to existing semiparametric estimators, proportional intensities is not assumed. The new estimators are asymptotically normally distributed. The estimator of b is root-n consistent. The estimator of h(y|x) achieves the one-dimensional rate of convergence. Thus the single-index assumption eliminates the "curse of dimensionality". The estimators perform well in Monte Carlo experiments. |
| Submitted paper full-text in .pdf |